Routing issue in MPLS VPN

Unanswered Question
Jan 30th, 2007
User Badges:

Problem context & Scenario description -

We have recently implemented MPLS IP VPN with AT&T. We have 3 main offices and 10 other branch offices world wide and they all are connecting to AT&T MPLS VPN either via frame-relay or ATM ePVC. AT&T routes various subnets of 10/8 like for boston office, for California office, for India office and similarly other /16 subnets for branch offices. We run BGP between our edge router and with AT&T edge router at each location. However, we have kept one IPLC line as a back up between India office and US Boston office. We want to automate the failover to this IPLC each time there is an issue on MPLS VPN from India. As you might be aware that recently there was a huge outage in Asia pac region due to Taiwan earthquake, at that time there were major routing issues in ATT network due to which our traffic from India office could not reach US offices via Asia PAC AS. Similar problem happened few days back also when AT&T ran into routing issues in their network and our India traffic to/from US could was lost inside the at&t network.

Problem is that during such routing issue at&t keep advertising our /16 subnets from their AS and those routes do not get evaporated from routing table as it happens in case of some physical link goes down and IGPs like OSPF or EIGRP recalculates the routes. As the old routes do not vanishes hence our router do not recalculates the new routes and traffic keeps taking the at&t path and getting lost in their AS. Every time such problem occurs we need to bring the back up iplc in BGP manually and only after that India-US traffic takes the IPLC path due to shortest AS distance. We want to avoid this manual configuration and want something automated. Please suggest the best strategy to achieve that in such scenario.

Is there any possibility routing decision can be made based upon ping or trace result??

Scenario diagram attached

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 3 (1 ratings)
lgijssel Tue, 01/30/2007 - 04:40
User Badges:
  • Red, 2250 points or more

You might be able to use object tracking for this case:

You could then automatically switch over when an ip adress becomes unreachable. However, I should be weary for the intermittent connectivity issues (high latencies, 20-60%loss) that are common for this kind of situations.



mheusinger Tue, 01/30/2007 - 04:55
User Badges:
  • Green, 3000 points or more


this is a copy and paste from the answer to your other identical posting:

the best bet would be "object tracking". This feature allows to trigger rerouting based on an "object". An object can be defined to be f.e. reachability to a server through ping. Basically you can instruct a router to ping an IP with a configurable frequency. If f.e. 3 consecutive pings are failing then rerouting through a backup can be triggered.

In fact there are much more options to define an object. A sample configuration would look like this:

ip sla monitor 1

type echo protocol ipIcmpEcho

timeout 2000

frequency 5

threshold 3

request-data-size 1490

ip sla monitor schedule 1 start-time now life forever


The full picture is given f.e. in "Reliable Static Routing Backup Using Object Tracking"

Hope this helps! Please use the rating system.

Regards, Martin

virenderj Wed, 01/31/2007 - 01:32
User Badges:

Hello Martin

thanks for your reply, this is actually something I was looking for and would definitly help.

However, there is little more complexity involved in my scenario. We redistribute lot of subnets from our hub offices like boston and california and there are tons of other networks in US. So static routing may not be that scalable in our case. We just want if there is any issue in india to US connectivity only the routes redistributed or advertised from US takes the IPLC back up path and india to UK traffic continute to take the normal mpls path from India. Please suggest. Net net if we can integrate some sort of dynamic routing with object tracking it would be great.



stephtchoko Wed, 01/31/2007 - 03:05
User Badges:
  • Bronze, 100 points or more

Hi all,

Tracking object use more cpu ?


mheusinger Wed, 01/31/2007 - 03:08
User Badges:
  • Green, 3000 points or more

Hi Virender,

another option would be to use "Optimized Edge Routing" (OER) which will inject routes in BGP based on reachability. Have a look at "Cisco IOS Optimized Edge Routing Configuration"

and especially the section with BGP: "Configuring iBGP Peering on the Border Routers"

Regards, Martin


This Discussion